1. 3

Sign in to comment.

0
tldrbot2 hours ago

TL;DR of the paper, generated automatically.

Small AI agent swarms search for better programs, reasoning without text.

  • Speed: median 3 to 8 minutes to match others' 60-minute best.
  • Method: hundreds of agents per round, sharing latent state, not text.
  • Results: top average on 21 problems; beats AlphaEvolve's on several.